Risk parity aims to balance assets' contributions to portfolio risk. That is different from giving each asset the same dollar weight. An inverse-volatility allocation is a simpler related rule: assets with lower measured volatility receive higher weights.
State the calculation precisely
For an illustrative model using SPY, TLT and GLD:
- At each monthly decision, measure each asset's volatility over the preceding 60 trading days.
- Set its raw weight to one divided by that volatility.
- Normalize the weights to sum to 100 percent.
- Specify how missing observations, zero volatility and execution timing are handled.
The 60-day lookback is an example parameter, not an optimal setting. Inverse-volatility weights ignore correlations and do not generally equalize portfolio risk contributions. A full equal-risk-contribution model needs an explicit covariance estimate and optimization method.
Inspect the example
The public card links to a changing model record. Inspect its run date, input schema, actual weights, history and warnings. Do not assume its current parameters match an example prompt or that a public model exposes its owner's code.
Leverage Aversion and Risk Parity, by Asness, Frazzini and Pedersen, provides background on risk-based allocation. Its research is separate from the current ENSEMBLE model and does not validate this particular implementation.
Test the choices that matter
Compare one lookback or rebalance schedule at a time. Inspect how weights respond to changing volatility, whether any asset becomes concentrated, and how much the modeled portfolio trades. Use identical dates and cost assumptions when comparing against a fixed allocation.
A low measured volatility is not a guarantee of low future risk. Correlations can change, assets can decline together, and estimates can react after a shock. Adding leverage introduces additional financing and loss risks; do not infer that those effects were measured in an unlevered backtest.
Build your own model
This prompt illustrates a specification. Review the generated code in your own model, verify data coverage and disclose any backfill assumptions. Use the portfolio walkthrough for the research and comparison steps. Saved inputs and runs make the choices inspectable, but do not guarantee immutable snapshots of every dependency.
ENSEMBLE is research software. Backtests are hypothetical and past performance does not guarantee future results.
Frequently asked questions
- Is inverse-volatility weighting the same as risk parity?
- Not generally. Inverse-volatility weights use each asset's volatility. Equal risk contribution also depends on correlations between assets; the two approaches coincide only under particular assumptions.
- Does this example use leverage?
- The illustrative specification is long-only and fully invested without borrowing. That does not describe every strategy marketed as risk parity.
- Which lookback is best?
- There is no universally best lookback. Choose a hypothesis, compare a small number of alternatives over matched dates, and inspect sensitivity and turnover.
Related
- StrategiesThe 60/40 portfolio: rules, variants, and a backtest
Define a stock-and-bond allocation, inspect a public 60/40 model, and test rebalancing or asset changes over matched dates. Learn what the model shows and what it leaves out.
- StrategiesDual momentum: the rules, the evidence, and a backtest
Define relative and absolute momentum separately, inspect a public model, and compare a parameter change. The example explains the research workflow without promising a historical or future result.
- BacktestingBacktest metrics explained: CAGR, volatility, Sharpe ratio, and max drawdown
Read ENSEMBLE backtest metrics with their actual conventions: CAGR, annualized volatility, a zero-risk-free Sharpe calculation, maximum drawdown and trading activity. Compare matched windows and inspect assumptions.
- BacktestingCommon backtesting mistakes and how to avoid them
The ways a backtest lies: lookahead bias, survivorship bias, overfitting, ignored costs, short or regime-limited samples, and misread metrics. What each one looks like, how to detect it, and what a backtesting tool can and cannot protect you from.